In the JournalismAI Collab, journalists from different news organisations came together to explore how AI-powered tools might improve our journalism. Six months of brainstorming and ideation gave us many innovative ideas to energise newsrooms. They also helped us understand the challenges of applying AI to journalism, such as the lack of AI literacy. Our experience in the Collab helped us understand how the roles in the newsroom might change to pave the way for profitable and impactful journalism through augmented data management.
Our team – with members in Germany, India, France, and Sweden – looked at the topic of connecting users to quality journalism with AI-powered summaries. But some of the most relevant lessons we took away from the Collab were not just about the technologies, but rather about the process of innovation and collaboration, and even the nature of journalism itself. This is what I learned while looking back with my team at our Collab experience:
Christina Elmer (Deputy Head of the Editorial RnD team at DER SPIEGEL) says: “The Collab is centred around technological innovations but has repeatedly led us to discussions about the core of our journalistic work. For me, this aspect was particularly impressive. What is the special value of our journalism that makes our products distinctive and ensures that people trust us? And in which areas can we safely automate formats using artificial intelligence, without putting trust at risk? To discuss such ethical and societal questions with like-minded fellows was especially inspiring and has connected us globally.“
She adds: “In order to assess the usability of AI systems, we had to analyse journalistic workflows, distinguish repetitive tasks from creative ones, and reflect on strategic issues. Artificial intelligence cannot replace journalists so we have to build up hybrid processes, interconnecting humans and AI. To succeed at that, we need a higher level of algorithmic literacy, a process-based understanding of our work, and completely new roles in the newsroom. Working in such an environment is, in my opinion, a fascinating vision – but we have to pass on this enthusiasm to all our colleagues to make this vision come true.“
Didier Orel (Head of Data Analytics at TX Group) reflected that the diversity of the Collab helped his team to keep in mind the diversity of potential interests and use cases as well. Didier believes in the importance of raising awareness in newsrooms about the potential, as well as the limits, of AI-powered technologies. Newsroom leaders should understand enough the power of AI to be able to identify the tools and their concrete usages that have the most potential to keep readers engaged with our digital products.
Olle Zachrison (Head of Digital News Development at Swedish Radio) said: “Living under the constant squeeze of limited resources, there are a lot of ideas that newsrooms would never prioritise if they have to produce them manually – but could very well consider them if they could be automated. One example is short, translated audio summaries: to have an AI-powered tool read a summary of the main news stories in other languages could open up our content to a much wider audience. As one of our tests showed, an automated chain of 1) transcribing a Swedish audio piece, 2) translating it to English, 3) and summarising it, worked remarkably well, on the condition that the piece was short, newsy and in one clear voice.“
While talking about the AI summarisation tool that our team tried to develop during the Collab, Olle adds: “Our collaborative exploration shows the importance of exposing journalists to the whole range of opportunities that can be found in the field of structured journalism. The mere knowledge that AI-powered tools can automatically extract audio summaries, key quotes or Q&A’s from any article thrills the creative imagination. Having this conversation with people from different disciplines, countries, and media types has been very fruitful as one small part of the industry’s quest to develop tomorrow’s journalistic formats.“
Uli Köppen (Head of the AI + Automation Lab at the German public broadcaster Bayerischer Rundfunk) said: “We aim for better journalism, with new tools and new skill sets. Translated to our Collab challenge, this meant to not focus only on using a new tool for summarisation but to reassess what kind of content is most representative to define our news brand, how to ideally raise attention for this type of content, and to re-think workflows where automated summarisation might add value.”
Cécile Schneider, who steers product development at the AI + Automation Lab of Bayerischer Rundfunk, talked about some of the challenges we encountered: “Improving the quality of automated summaries requires a lot of fine-tuning towards very specific input and output formats. Our collaboration with summarisation start-up Agolo showed us what this could look like through their impressive experience of working with the Associated Press.“
Cécile also discussed the monetisation component of our research: “It will be interesting to follow commercial solutions in this field and see which journalistic use cases they can monetise, given the mostly-stretched media budgets. Our experiment showed that needs around summarisation may vary greatly between newsrooms, and also the availability of suitable training data for non-English languages can be an issue.“
We all agreed on the importance of spreading AI literacy in our newsrooms and helping colleagues understand the growing importance of these new technologies in augmenting newsroom workflows. Exploring new AI tools also gives newsrooms the opportunity to re-imagine those same workflows and how they can be improved with a data-based approach.
“In our case, testing summaries as teasers brought us to discussing other formats of link-teasers such as quotes and questions. These might be more captivating for users and might deserve further exploration – as the results of the A/B testing at Der Spiegel show in our study“, Uli added.
In our JournalismAI collaboration, the focus of our combined brainstorming took us from first principles to future thinking. From spreading AI literacy to using AI-powered summarising tools, to AB testing and putting together an extensive final report, it was a great and insightful journey. And, it’s not going to end here of course.
This piece was written by Pratyush Ranjan, Senior Editor at Jagran New Media in India and one of the members of ‘Team 3’ of the JournalismAI Collab. It sums up their main lessons learned from working together in 2020 to explore how automated summaries can be used by newsrooms to make their content more discoverable and accessible. Read the findings of their study on our website.
The JournalismAI Collab is a project of POLIS, supported by the Google News Initiative.